Q550: Individual Projects

Each student will select a domain that is of personal interest, and will propose a simple experiment and a simple model. The model should not be complex—it must be simple enough that the student can clearly explain it to the class and be confident that it is accurately programmed. The model should also have a clear interpretation in terms of cognitive processes; purely "descriptive" models do not qualify. The proposal must be approved by the instructor.

The purpose of this project is to produce something you can eventually present at a conference or follow through to a publication. Note, however, that you cannot use the data collected in this class to present anywhere but in this class (IRB rules). We can use our experiments as pilot studies, and I would be happy to help you pursue any of these projects after the class.

The experiment must be simple enough to run in about 20 minutes. We will be using a computer lab with each of us doing experiments side-by-side, so avoid any fancy auditory stimuli if you can. If you need to run fewer individuals in isolated booths, let me know and we may move to my lab for one data collection session.

There are (at least) three approaches to the individual project:

1) identify an existing model and the phenomena it explains, simplify the model to the basic mechanisms necessary to address the core phenomenon, and fit the model to the collected data.

2) identify two competing models of a phenomenon, code the models, and create an experiment which will allow you to constrain between the models qualitatively or quantitatively.

3) identify a robust phenomenon for which there is currently no explanatory model, and create/fit one.

Students will program a model and collect/analyze experimental data (using the other students and instructor as subjects). Each student will present two 15-minute talks: the first one describing the empirical phenomenon and theoretical background, and the second one proposing the simple model and experiment. At the end of the course, each student will submit two final papers (one small, one larger).

Components

1) Half-page Proposal (Due Feb. 1)

o Briefly summarize the domain, and the empirical phenomenon of interest

o Briefly describe your proposed modeling approach (above 3 options)

o Briefly describe the type of experiment you might conduct to demonstrate the phenomenon

o Cite a couple of core papers on your topic, so we might read them before your presentations

Note: This is a very brief proposal of your direction. You are not expected to know all the specifics of the model/experiment, and this info may change as you go along. It just sets a beginning roadmap for your project.

2) Talk 1: Background and Empirical Phenomenon (15 minutes)

o Present some background on the domain of interest for those of us unfamiliar, and why we should care

o Describe the specific empirical phenomenon of interest

o Describe the types of experiments that are used to demonstrate the phenomenon

o Describe your modeling approach and general experiment

3) Talk 2: Specifics of Model and Experiment (15 minutes)

o Briefly remind us of your domain, empirical phenomenon, and why we should care about it

o Describe the specifics of your model:

> Representation

> Process

> Parameters and their interpretation

> Output

o Describe the specifics of your experiment:

> Procedure

> Variables (IV, DV, data measurement scale)

> Proposed data analysis and model fit measures

o A priori predictions that your model(s) make(s) for the experiment

4) Experiment

o Keep it simple, with clear directions

o Do several dry runs to ensure you won't have a crash

o Humans do funny things we don't anticipate: Remember good programming and I/O bulletproofing

o The experiment should be a scaled-down version; it should only take 20 minutes from start to finish

5) Papers and Code

Hand in papers (2) and your model/experiment code by 4pm on May 4^th. I am not picky on the programming language you choose, nor do the model and experiment have to be written in the same language (some are better for matrix manipulation, some better for user I/O). Whatever works best for you (this statement is subject to software availability in the computer lab).

You will submit two papers. The major paper should be written first. The minor paper is simply a scaled-down version of the major paper, written as if you're submitting it as a paper to the Proceedings of the Cognitive Science Society. For the minor paper, as with proceedings papers, you are limited to 6 pages using the Society template. You can find the Word template and sample papers here. This minor paper is your opportunity to summarize the salient points of the work; you are welcome to borrow heavily from the major paper, as long as you can summarize your background, model, experiment, and findings in 6 pages. Proceedings from previous years can be found here.

The major paper should follow APA format, and should not exceed 15 pages (double spaced). Start with an introduction to the domain and phenomenon, why it is important, and brief description of relevant models (if any). You should also have some clear hypothesis you are testing. Then:

o Describe your model, representation, parameters, and predictions it makes for the task

o Describe the experiment, including the standard subsections to the Method: Subjects, stimuli, procedure

o Results: Present descriptive and inferential statistics to demonstrate your effect, and model fit or comparisons if applicable.

o General Discussion: Relate back to your introduction, and draw a conclusion about your hypothesis based on your data and model fitting. Suggestions for model modifications and future experiments/model comparisons.

CLICK HERE for an example of a simple experiment and model of information integration from Prof. Kruschke's previous version of this course.